Energy storage systems, particularly batteries, are essential in modern energy infrastructure. They power devices such as portable electronics, electric vehicles, and grid storage systems. A key aspect of managing these systems is accurately assessing their State of Health (SOH), which evaluates a battery's condition and performance compared to its original state.1
Image Credit: Gorodenkoff/Shutterstock.com
Accurate SOH estimation is important for predicting battery lifespan, ensuring safety, and optimizing performance. SOH is a key metric in energy storage systems that reflects a battery's condition and performance capabilities.
Traditional methods often lack the precision needed for real-time monitoring, which can result in unexpected failures and higher maintenance costs.2 Artificial Intelligence (AI)-based methods have introduced more reliable techniques for SOH estimation, addressing these challenges.
SOH is influenced by multiple parameters, including capacity, internal resistance, power fade, and cycle life, all of which undergo gradual degradation due to aging processes. These degradation mechanisms can be categorized into two primary types: cycle aging, which occurs as a result of charge-discharge cycles, and calendar aging, which progresses during storage periods regardless of usage.3
Managing factors such as temperature can help mitigate these effects and extend battery life.
AI's Role in SOH Estimation
Machine Learning Models for Estimating SOH
Machine learning algorithms are effective in improving the accuracy of SOH estimation for batteries. Techniques such as linear regression, support vector machines (SVM), and random forests are commonly used to model the relationships between battery parameters and SOH. More advanced methods, including artificial neural networks (ANNs) and deep learning, enhance accuracy by capturing complex nonlinear patterns in the data.2
SVMs are versatile regression tools that are particularly effective for datasets with non-linear relationships. They predict SOH using features such as voltage, current, temperature, and cycle count. Their performance depends on effective feature selection and hyperparameter tuning, including kernel adjustments.4
ANNs employ multi-layered models to learn the relationships between historical battery data and SOH values, iteratively refining weights and biases for accurate predictions. Their efficacy depends on careful architecture design, activation functions, and preprocessing.5
Deep learning, a specialized form of ANN, uses multiple interconnected layers to analyze complex data patterns, making it adept at capturing intricate battery behaviors.6 This ability makes it particularly suitable for real-world SOH estimation and proactive maintenance planning.
While these methods achieve high accuracy, success depends on sufficient training data and rigorous preprocessing, feature selection, and model optimization. These data-driven approaches enable precise SOH estimation, facilitating proactive maintenance and extending battery lifespan.
Real-Time Data Analysis for Accurate Predictions
Real-time data analysis is critical for ensuring accurate predictions in battery management systems, but it poses challenges due to the computational burden of complex AI algorithms. Simplified approaches are often necessary to balance accuracy and efficiency.
ANNs are a practical solution. They offer a straightforward training process, flexibility for data training, and low computational costs, making them suitable for critical real-time applications like electric power systems.3
Researchers have developed methods to optimize real-time performance. For instance, Khumprom et al. implemented a simple approach that uses real-time operational data, making the proposed model suitable for real-time applications.7
In other research, Li et al. proposed using SVM to develop a battery degradation model and accurately predict battery cycle numbers with reduced computational time.8 These advancements demonstrate how AI-driven methods optimize real-time data analysis, ensuring efficient and precise predictions in energy storage applications.
Benefits of AI in SOH Estimation
Improved SOH Accuracy
Hybrid algorithms, co-estimation methods, and intelligent models can significantly enhance the accuracy of SOH estimation. Combining multiple approaches increases precision and reduces the computational burden.
For example, Li et al. introduced the improved bird swarm algorithm optimization least squares support vector machine (IBSA-LSSVM) model for estimating the remaining useful life (RUL) of lithium-ion batteries. The model achieved a root mean square error of 0.01, demonstrating higher prediction accuracy compared to other models.9
Transfer learning (TL) has also been used to reduce computational demands while maintaining accuracy. A 2021 study introduced a framework combining TL and network pruning to develop compressed convolutional neural network (CNN) models that operate on small datasets with improved accuracy and reduced complexity.10
Several companies are at the forefront of integrating AI into energy storage management. For example, Stem Inc. utilizes AI-driven software to optimize energy storage and management, enhancing the efficiency and reliability of energy systems. Similarly, Form Energy is developing innovative energy storage solutions that incorporate AI to improve performance and longevity.
Early Detection of Issues and Reduced Maintenance Costs
AI-driven approaches in energy storage systems enable early detection of potential issues, preventing unexpected failures and minimizing downtime. By continuously monitoring parameters such as voltage, current, temperature, and cycle count, AI models can identify anomalies and predict degradation trends before they lead to critical problems.3
For instance, advanced machine learning techniques, such as ANNs and SVMs, can forecast RUL and detect irregularities in real time. Early issue identification enhances system reliability and optimizes maintenance schedules, cutting down unnecessary servicing costs.3
Predictive maintenance enabled by AI also reduces reliance on manual inspections, lowering labor expenses and operational disruptions. Ultimately, these capabilities contribute to more cost-efficient and reliable energy storage systems.
Challenges in Implementing AI
Quality and Availability of Data
High-quality and accessible datasets are essential for accurate SOH estimation in energy storage systems. Organizations like NASA, CALCE, and Oxford University provide valuable datasets for training and validating intelligent algorithms, covering diverse conditions such as temperature variations and long-term cycling. Specialized datasets, like the eVTOL dataset, focus on real-world applications, improving predictive algorithms for electric aircraft.3
Tang et al. developed a migration-based machine learning approach that combines industrial data with accelerated aging tests to generate high-fidelity battery aging datasets. This method reduces experimental time by up to 90 % while maintaining an error margin of less than 1 %.11
In addition to ensuring high-quality and extensive datasets, extracting key features from large datasets for model training remains a significant technical challenge that must be addressed to enhance predictive accuracy and model reliability.
Complexity of AI Models
The complexity of AI models in energy storage systems introduces several challenges. Advanced models, such as deep learning and hybrid approaches, often require significant computational resources, making real-time implementation difficult. These models also demand large, high-quality datasets for training, along with extensive preprocessing and feature selection.6
Hyperparameter tuning and optimization further complicate the deployment process, while the black-box nature of many AI models raises concerns about interpretability, particularly in safety-critical applications.3 Overcoming these challenges requires balancing model sophistication with efficiency and developing more interpretable and resource-efficient AI solutions.
AI and the Future of Battery Health Management
AI is significantly improving the estimation of SOH in energy storage systems. By utilizing machine learning models and real-time data analysis, AI enables more accurate and timely assessments of battery health. This contributes to enhanced performance, increased lifespan, and lower maintenance costs.
While challenges persist regarding data quality and model complexity, ongoing research and industry applications are advancing the integration of AI in energy storage management. These developments are contributing to the creation of more reliable and efficient energy solutions.
Advancing Safer Battery Solutions with South 8 Technologies
References and Further Reading
1. Sarmah, S. B.; Kalita, P.; Garg, A.; Niu, X.-d.; Zhang, X.-W.; Peng, X.; Bhattacharjee, D. (2019). A Review of State of Health Estimation of Energy Storage Systems: Challenges and Possible Solutions for Futuristic Applications of Li-Ion Battery Packs in Electric Vehicles. Journal of Electrochemical Energy Conversion and Storage. https://asmedigitalcollection.asme.org/electrochemical/article/16/4/040801/726442/A-Review-of-State-of-Health-Estimation-of-Energy
2. Demirci, O.; Taskin, S.; Schaltz, E.; Demirci, B. A. (2024). Review of Battery State Estimation Methods for Electric Vehicles-Part Ii: Soh Estimation. Journal of Energy Storage. https://vbn.aau.dk/en/publications/review-of-battery-state-estimation-methods-for-electric-vehicles--2
3. Raoofi, T.; Yildiz, M. (2023). Comprehensive Review of Battery State Estimation Strategies Using Machine Learning for Battery Management Systems of Aircraft Propulsion Batteries. Journal of Energy Storage. https://www.sciencedirect.com/science/article/abs/pii/S2352152X22024756?via%3Dihub
4. Zhang, Y.; Liu, Y.; Wang, J.; Zhang, T. (2022). State-of-Health Estimation for Lithium-Ion Batteries by Combining Model-Based Incremental Capacity Analysis with Support Vector Regression. Energy. https://ideas.repec.org/a/eee/energy/v239y2022ipbs0360544221022349.html
5. Luo, Y.-F.; Lu, K.-Y. (2022). An Online State of Health Estimation Technique for Lithium-Ion Battery Using Artificial Neural Network and Linear Interpolation. Journal of Energy Storage. https://www.sciencedirect.com/science/article/abs/pii/S2352152X22010647
6. Fan, Y.; Xiao, F.; Li, C.; Yang, G.; Tang, X. (2020). A Novel Deep Learning Framework for State of Health Estimation of Lithium-Ion Battery. Journal of Energy Storage. https://www.sciencedirect.com/science/article/abs/pii/S2352152X20315784?via%3Dihub
7. Khumprom, P.; Yodo, N. (2019). A Data-Driven Predictive Prognostic Model for Lithium-Ion Batteries Based on a Deep Learning Algorithm. Energies.
8. Li, X.; Shu, X.; Shen, J.; Xiao, R.; Yan, W.; Chen, Z. (2017). An on-Board Remaining Useful Life Estimation Algorithm for Lithium-Ion Batteries of Electric Vehicles. Energies. https://www.mdpi.com/1996-1073/10/5/691
9. Li, L.-L.; Liu, Z.-F.; Tseng, M.-L.; Chiu, A. S. (2019). Enhancing the Lithium-Ion Battery Life Predictability Using a Hybrid Method. Applied Soft Computing. https://www.sciencedirect.com/science/article/abs/pii/S1568494618305714?via%3Dihub
10. Li, Y.; Li, K.; Liu, X.; Wang, Y.; Zhang, L. (2021). Lithium-Ion Battery Capacity Estimation—a Pruned Convolutional Neural Network Approach Assisted with Transfer Learning. Applied Energy. https://www.sciencedirect.com/science/article/abs/pii/S0306261920317773
11. Tang, X.; Liu, K.; Li, K.; Widanage, W. D.; Kendrick, E.; Gao, F. (2021). Recovering Large-Scale Battery Aging Dataset with Machine Learning. Patterns. https://www.sciencedirect.com/science/article/pii/S2666389921001458
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.